Paris, 15 October 2026. Embedded France and six partner organisations - the Eclipse Foundation, CEA, Leminnov, Quantyss, the European Trustworthy AI Association and Lianeo - today release the English edition of the Structured Glossary of Embedded AI. Produced by Embedded France’s Embedded AI Working Group, led by Eliane Fourgeau, CEO of Quantyss, and first published in French in January 2026, the glossary is now available in both languages. It can be read online or downloaded as a PDF, under a Creative Commons licence, from a platform hosted by the Eclipse Foundation.
The need for a shared vocabulary
Artificial intelligence is now everywhere, in vehicles, trains and aircraft, in medicine, in industrial equipment, in energy grids, to name only the most familiar.
But unlike non-embedded AI, embedded AI operates under strict physical constraints: limited memory, latency constraints, demonstrable safety. These constraints sit at the crossroads of multiple disciplines, blending the expertise of data and AI scientists, software engineers, hardware designers, systems architects, cybersecurity, safety and compliance specialists. Every profession has its own jargon, habits and routines, and when people do not speak the same language, the risk of misunderstanding is high.
These gaps are not a matter of style. They cost real money: ambiguous specifications, misread requirements, validation campaigns that have to be run again, certification files that keep growing. In an embedded system, a misunderstood word can turn into a flawed architectural decision, and such decisions are not easily undone.
Something had to be done, and quickly, to help these worlds understand one another. That is exactly what Embedded France’s Embedded AI Working Group set out to do.
A structured reference of more than 300 definitions
When the working group was set up in 2024, its starting point was simple: no real collaboration is possible unless everyone can describe each task and each situation precisely, in terms that every stakeholder understands. Nearly a year of work produced a first French-language glossary, presented in January 2026 at the 18th Assises de l’Embarqué.
Today the group is unveiling a bilingual edition. It brings together more than 300 definitions organised in 13 chapters that follow the life cycle of a system: data and acquisition, models and their optimisation, software and tooling, architecture and hardware, emerging paradigms, testing, validation and certification, trustworthy AI, governance and life cycle, regulation and standards, and sector applications.
Each term sits within its chapter and is linked to related notions, so readers find not only a definition but where it fits within a system. It is a genuine working guide.
The glossary covers the frameworks industry now has to reconcile, from the ISO and DO standards that apply to critical systems through to the AI Act and the Cyber Resilience Act, and includes the field’s newest concepts: generative AI, agentic AI and physical AI.
A common base, open to all
The glossary was started within Embedded France’s Embedded AI Working Group, with the Eclipse Foundation, CEA, Leminnov, Quantyss, the European Trustworthy AI Association and Lianeo, each bringing a complementary perspective, from research and industry to governance and open source. They compared how they use each term and, working collegially, agreed on a first set of common definitions. The glossary does not claim to be exhaustive or prescriptive: it is a common base, open to corrections and contributions from everyone who uses it. “This glossary provides European industry with a shared foundation to master embedded and physical AI. By structuring our vocabulary, we structure our skills,” said Eliane Fourgeau, Vice-President of Embedded France. “It is a strategic tool to strengthen technological sovereignty, accelerate talent development and support the emerging physical AI economy across all sectors.”
Openly published under a Creative Commons licence
Released under the Creative Commons Attribution 4.0 International licence, the glossary can be read, downloaded, reproduced, translated and adapted freely, including for commercial use, on one condition: credit the source. Companies can attach it to a specification, schools can build courses around it, and standards bodies can cite it.
It is hosted by the Eclipse Foundation within the Eclipse Aidge project and maintained as an open project: corrections, clarifications and new terms are submitted through its public repository.
This new release of the glossary will be presented at the Assises de l’Embarqué on 26 January 2027, an edition devoted to physical AI.
“A shared vocabulary is an important foundation for collaboration, particularly when AI safety, security, and reliability are at stake,” said Mike Milinkovich, Executive Director of the Eclipse Foundation. “Making this glossary openly available gives developers, researchers and industry a common reference point, helping them work together more effectively to advance trustworthy embedded AI.”
THE GLOSSARY AT A GLANCE
|
• More than 300 definitions across 13 chapters • Seven initial contributors, open to everyone’s contributions • Creative Commons Attribution 4.0 International licence • Read the glossary: https://pages.eclipse.dev/eclipse/aidge/embedded-ai-glossary/ • Contribute: https://gitlab.eclipse.org/eclipse/aidge/embedded-ai-glossary |
Embedded France
Embedded France is the French professional organisation dedicated to embedded technologies. It brings together industrial, academic and institutional stakeholders involved in embedded systems, critical software, hardware architectures and embedded AI solutions. Embedded France leads sector-specific working groups, contributes to national and European initiatives, and supports ecosystem structuring to reinforce technological sovereignty and talent development.
Eclipse Foundation
The Eclipse Foundation is an international non-profit organisation providing a vendor-neutral home for open source innovation, with more than 450 projects spanning a broad range of technology domains and over 300 member organisations worldwide. Its Sovereign AI Foundation brings members together to strengthen choice and control over AI technologies. The Foundation's growing AI ecosystem includes projects such as Eclipse Aidge, Eclipse Theia AI, and Eclipse LMOS, spanning embedded and industrial AI, AI-assisted development, and agentic AI. Together, these efforts help organisations develop and adopt open AI technologies with greater transparency, flexibility, and control.
CEA
The CEA works to promote the scientific, technological and industrial sovereignty of France and Europe in four key areas: low-carbon energy, digital technology, the medicine of the future, and defence and security, drawing on excellence in fundamental research. As a public technological research organisation, the CEA contributes to the scientific, technological and industrial sovereignty of France and Europe. For further information: www.cea.fr
Leminnov
Leminnov is a French deep-tech company engineering the trusted digital infrastructure that brings Cloud Native, Edge Computing and AI together. It designs and industrialises secure, resilient and interoperable platforms that enable organisations to deploy and operate AI and distributed workloads from cloud to edge, leveraging open technologies, Platform Engineering and DevSecOps practices. As a member of Embedded France and Numeum, and part of the French Tech ecosystem, Leminnov contributes to building open and trusted technology ecosystems that strengthen interoperability, industrial innovation and European technological sovereignty.
Quantyss
Quantyss is a strategy consultancy dedicated to deep‑tech growth. It guides startups, SMEs and mid‑caps from vision to market traction - aligning technology, investors and industry.
European Trustworthy AI Association
The European Trustworthy AI Association is a non-profit organisation established in 2025 in Brussels, Belgium, by industrial leaders, building on the legacy of the Confiance.ai programme. It brings together some thirty European organisations, industrial companies, startups, academic institutions and research centres, with the mission of empowering industry with state-of-the-art open-source methods and tools to engineer AI-based systems that can be trusted and comply with regulations.
Lianeo
Lianeo is a consulting company specialising in embedded systems and IoT. Since 2012, it has supported manufacturers in telecoms, automotive, medical, defence and aerospace in developing their embedded and critical software.
PRESS CONTACTCendrine Barruyer, General Delegate, Embedded France |